Marshall Ward

MOM6 does two-space indents.  My personal preference is 4 spaces since 2-space indent "errors" are harder to detect visually and can slip in more easily.  But I haven't noticed much...

I would like to see greater support for bit-reproducible numerical operations. This is a very high priority for us since our models are used in weather and climate forecasting, and...

That's right, we are primarily focused on reproducing the bits over a range of configurations, and at least for now we're not too concerned with the overall accuracy of the...

The MPI summation method uses an extended fixed precision method, which basically converts the floats to tuples of integers, each value denoting a fixed precision over a selected range. MPI...

The underlying issue is that if we invoke a transcendental intrinsic, then we really just don't know where it is coming from. For example, we once had our system math...

@certik True, though it's not necessarily accuracy that matters in our case. I just want the same numbers!

@milancurcic You've captured the main issue, which is an implicit link to some `libm.so.*` binary (or a custom library, I suspect, in other cases). We have had situations where there...

@warrickball That is encouraging news. I also supported the idea of testing with a prescribed `libm`. However, there is the deeper problem that the compiler is under no obligation to...

Python 3 formatting is becoming quite widespread; I know that C# and Rust use them. It would be great to reach this point, but I agree with @cmacmackin that it...

I agree that the reference implementation should use Fortran as much as possible. If there are cases where it cannot be used, then it will highlight a critical limitation of...